My question is about the direct sparse solver routine. I am using the dss interface to facotrize and invert a large coefficient matrix.I am successfully inverting the orginal problem Ax = b. In my program I have this line:
error = DSS_SOLVE_COMPLEX(handle_b, MKL_DSS_DEFAULTS, rhs_hb, 1, solution )
Also I need to solve the conjugate transpose version of this system of equations (AHx = b).So i have following line:
error = DSS_SOLVE_COMPLEX(handle_c, MKL_DSS_CONJUGATE_SOLVE, xdzxy, 1, intersolx)
The reference manual states: "MKL_DSS_CONJUGATE_SOLVE option added to the parameter opt enables solving a conjugate transposed system AHx = b based on the factorization of the matrix A. This option is equivalent to the parameter iparm(12)= 1 in PARDISO."
When I look in the manual of the PARDISO project: http://www.pardiso-project.org/manual/manual.pdf.
There is no such option to solve the conjugate transposed system in iparam(12). It would be nice if someone can help me and explain why in the MKL version of the PARDISO solver has this option.
You are referring to the manual of the PARDISO solver developed by Olaf Schenk et al., whereas the documentation of MKL_DSS_CONJUGATE_SOLVE is refering to the Intel MKL PARDISO. While they share some common characteristics, some features are available in only one of them. You can check this link http://software.intel.com/sites/products/documentation/hpc/mkl/mklman/index.htm#GUID-431916D5-B76D-4... for a documentation of the Intel MKL PARDISO.